Detecting differential gene expression in subgroups of a disease population

Area of Science:

• Biostatistics
• Genomics
• Medical Research

Background:

• Disease populations often comprise subgroups with unique genetic or phenotypic variations.
• Identifying these variations is crucial for understanding disease mechanisms and developing targeted therapies.
• Mixture distributions are expected when analyzing variables in disease groups with subgroups.

Purpose of the Study:

• To evaluate statistical methods for identifying differential expression in disease subgroups.
• To highlight the drawbacks of approaches that ignore standard statistical theory.
• To compare theoretically derived methods with an ad hoc outlier identification approach.

Main Methods:

• Comparison of theoretically motivated statistical tests with an outlier-sum statistic.
• Analysis of differential gene expression in simulated disease populations with subgroups.
• Evaluation of statistical power, false positive rates, and robustness.

Main Results:

• The outlier-sum statistic offers minimal advantage over a standard t-test, even in ideal conditions.
• The outlier-sum statistic exhibits poor robustness, difficulty in calibration, and high false positive rates.
• The asymmetric treatment of groups in the outlier-sum statistic limits its discriminatory ability.

Conclusions:

• Ad hoc methods like the outlier-sum statistic can be unreliable for detecting differential expression in disease subgroups.
• Adherence to sound statistical theory is essential for developing robust and accurate methods in genetic association studies.
• The t-test remains a more reliable approach than the outlier-sum statistic for identifying differential expression in many disease settings.